feat(identity): 建档与刷新写别名按保留规则标语言，不再写空语言的 other

Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01Cxwgnn2pbzQ7rXv3rq7fei

apps/server/src/modules/sources-identity/service/entities.ts

@@ -1,4 +1,4 @@
-import type { MetadataCandidate } from "@umt/adapters";
+import { allowedLanguage, type MetadataCandidate, textMatchesLanguage } from "@umt/adapters";
 import type { AliasKind, AliasSource, EvidenceLevel, MediaKind } from "@umt/contracts";
 import { sql } from "@umt/db";
 import { normalizeTitle, type ParsedRelease } from "@umt/domain";
@@ -171,6 +171,11 @@ export async function createEntityFromTmdb(
   return { entityId: entity.id, created: true };
 }
 
+/**
+ * TMDB 详情登记别名(来源 tmdb)。多语言保留规则(adapters language-policy.ts):details() 已只留简中 / 繁中 / 英文 /
+ * 原语言的本地化名、译名与别名(别名带按地区 + 文字推断的语言),这里按语言写入,不再写空语言的 other。
+ * 本地化名的文字与语言不符时(TMDB 缺该语言译名时回退成原名)不按该语言登记。
+ */
 export async function seedAliases(exec: Exec, entityId: string, m: MetadataCandidate): Promise<void> {
   // 同一文本只登记一次(按首个出现的类别):TMDB 的译名/别名列表常重复主标题与原名
   const seen = new Set<string>();
@@ -180,12 +185,21 @@ export async function seedAliases(exec: Exec, entityId: string, m: MetadataCandi
     seen.add(n);
     await addAlias(exec, entityId, text, kind, "tmdb", lang);
   };
-  await add(m.title, "translated", null);
+  const work = { originalLanguage: m.originalLanguage, originCountry: m.originCountry };
   if (m.originalTitle) await add(m.originalTitle, "original", m.originalLanguage);
+  let n = 0;
   for (const t of m.titles) {
-    if (t.kind === "localized") await add(t.text, "translated", t.lang);
+    if (t.kind === "original" || !t.lang || n >= 50) continue;
+    const lang = allowedLanguage(t.lang, work);
+    if (!lang || !textMatchesLanguage(lang, t.text)) continue;
+    const kind: AliasKind = t.kind === "alias" ? (lang.endsWith("-Latn") ? "romanized" : "other") : "translated";
+    await add(t.text, kind, lang);
+    n++;
   }
-  for (const a of m.aliases.slice(0, 50)) await add(a, "other", null);
+  // 主标题(建档时可能来自语言链):上面没登记到时补一条,语言按同文本的本地化名(文字相符才标)
+  const loc = m.titles.find((t) => t.text === m.title && t.kind === "localized" && t.lang);
+  const titleLang = loc?.lang && textMatchesLanguage(loc.lang, m.title) ? allowedLanguage(loc.lang, work) : null;
+  await add(m.title, "translated", titleLang);
 }
 
 export class IdentityConflictError extends Error {

apps/server/src/modules/sources-identity/service/metadata-rich.ts

@@ -1,9 +1,9 @@
 import {
   DEFAULT_METADATA_LANGUAGES,
-  scriptMatches,
   type TmdbRichDetail,
   type TmdbRichEpisode,
   type TmdbRichSeasonDetail,
+  textMatchesLanguage,
 } from "@umt/adapters";
 import type {
   EntityMetadataField,
@@ -25,6 +25,7 @@ import { ensureSeason, promoteUnits } from "./metadata-totals.ts";
 // - 作品:TMDB 富详情(fetchRichDetail)按设置 metadata.languages 取标题 / 简介 / 标语(MD-2:简介 / 标语缺省只回退到中文变体),另存中文名 titleZh、类型、评分、
 //   状态、时长、季数、主创、外部 ID;来源值写 metadata_field(language = 实际取到的语言),同字段换语言时删旧行;
 // - 别名:translations / alternative_titles 写 title_alias(来源 tmdb,标语言);已有同文本别名只补空的语言;
+//   只写简中 / 繁中 / 英文 / 原语言(多语言保留规则,adapters language-policy.ts);
 // - 季 / 集:fetchRichSeason 的每集 name / overview / airDate / runtime / stillPath / vote 写 metadata_field(content_unit_id),
 //   投影到 content_unit.metadata(字段级,可锁),name → title 列、airDate → air_date 列;只写变化的字段,
 //   不改编号、不碰路径绑定与资产覆盖。集单元不存在时按 TMDB 建(confirmed,挂季单元)。
@@ -152,26 +153,22 @@ export async function writeRichEntityFields(
   );
 }
 
-/** 别名的地区 → 语言:只在文字相符时标(CN 地区的英文别名不标 zh-CN) */
-function altLanguage(region: string | null, text: string): string | null {
-  if (!region) return null;
-  if (["CN", "SG", "TW", "HK", "MO"].includes(region)) return scriptMatches("zh", text) ? `zh-${region}` : null;
-  if (region === "JP") return scriptMatches("ja", text) ? "ja-JP" : null;
-  if (region === "KR") return scriptMatches("ko", text) ? "ko-KR" : null;
-  if (["US", "GB", "CA", "AU", "NZ", "IE"].includes(region)) return scriptMatches("en", text) ? `en-${region}` : null;
-  return null;
-}
-
 /**
- * translations(译名,kind translated)与 alternative_titles(别名,kind other)写 title_alias,来源 tmdb、标语言。
+ * translations(译名,kind translated)与 alternative_titles(别名,kind other;原产地区罗马字 romanized)写 title_alias,
+ * 来源 tmdb、标语言。多语言保留规则:fetchRichDetail 已只留简中 / 繁中 / 英文 / 原语言的译名,别名带按地区 + 文字推断的语言
+ * (推断不出的已丢弃);这里再要求译名文字与语言相符。
  * 同作品已有同文本别名时不再新增,只给来源 tmdb 且语言为空的那行补语言。最多 80 条。
  */
 export async function tagTitleAliases(exec: Exec, entityId: string, rich: TmdbRichDetail): Promise<number> {
-  const want: { text: string; kind: "translated" | "other"; language: string | null }[] = [];
+  type Kind = "translated" | "other" | "romanized";
+  const want: { text: string; kind: Kind; language: string | null }[] = [];
   if (rich.titleZh) want.push({ text: rich.titleZh, kind: "translated", language: rich.titleZhLanguage });
-  for (const t of rich.translations) if (t.title) want.push({ text: t.title, kind: "translated", language: t.lang });
+  for (const t of rich.translations)
+    if (t.title && textMatchesLanguage(t.lang, t.title))
+      want.push({ text: t.title, kind: "translated", language: t.lang });
   for (const a of rich.alternativeTitles)
-    want.push({ text: a.title, kind: "other", language: altLanguage(a.region, a.title) });
+    if (a.language)
+      want.push({ text: a.title, kind: a.language.endsWith("-Latn") ? "romanized" : "other", language: a.language });
   const rows = await exec
     .selectFrom("title_alias")
     .select(["id", "normalized", "language", "source"])

apps/server/src/modules/sources-identity/service/metadata.ts

@@ -361,7 +361,14 @@ export async function refreshMetadata(
     creatable = new Set(opts.full ? want : tracked);
     for (const n of want) {
       try {
-        richSeasons.push(await fetchRichSeason(tmdb, ident.external_id, n, langs, { now: rt.now(), ...textOpts }));
+        richSeasons.push(
+          await fetchRichSeason(tmdb, ident.external_id, n, langs, {
+            now: rt.now(),
+            ...textOpts,
+            // 语言链按作品原语言收敛(多语言保留规则)
+            originalLanguage: richDetail?.originalLanguage ?? detail.originalLanguage,
+          }),
+        );
       } catch (e) {
         ctx.logger.debug({ entityId: id, season: n, err: { message: (e as Error).message } }, "季详情获取失败");
       }